A mathematical model for biological motor learning based on synaptic dynamics

Abstract

We present a mathematical model to simulate the generation process of motor learning rate coefficients in human brain, focusing on neural communication through synaptic dynamics. The model consists of four major compartments, including three intraneuronal and one perceptron component. The motor learning rate coefficient is generated through biologically plausible changes in synaptic connections within the neural network, governed by the biological energy efficiency constraint. Our simulations for the first time demonstrated that the optimal motor learning rate coefficient as reported in previous research is biologically possible. We further validated the consistency, stability and robustness of the model. Additionally, we observed a distinct difference in the neural networks' structures between successful and failed motor learning processes. To achieve successful motor learning, it is essential that the interneuronal network is composed predominantly by inhibitory synaptic connections, and connections to the perceptron are primarily excitatory.